Very short‐term prediction model for photovoltaic power based on improving the total sky cloud image recognition

Due to the cloud cover, the power generation of photovoltaic power plant will reduce suddenly, that may lead to instability of the grid and bring some risks. Cloud observation is often used in the very shortterm prediction of photovoltaic power. However, in the haze weather, The quality of the total sky image will fall, the cloud recognition effect is reduced, further leads to decreased levels of photovoltaic power prediction.This study presents a photovoltaic power prediction algorithm,which takes into account the image quality decline caused by haze. So the algorithm improves the precision of photovoltaic power generation. Firstly, the extraction method using discrete Fourier transform checks whether clouds exist or not, and computing in the position of the solar facula, and then fix the facula. Next, using piece wise linear transformation algorithm for cloud enhancement, and then generate intensity layered cloud stain chart. Finally, the relationship between cloud and radiation is used to realise the very short‐term prediction of photovoltaic power. The experimental results show that the algorithm has a very good universality, and the cloud image can be identified by the fog and haze, therefore it improves the accuracy of the ultra‐short term prediction of photovoltaic power.